Across 6 shared benchmarks, Claude 3.5 Sonnet scores higher on 5 and Granite 4.2 3B on 1. The widest gap is LiveCodeBench v6, where Granite 4.2 3B scores 69.7 against 37.2.
| Benchmark | Claude 3.5 Sonnet | Granite 4.2 3B |
|---|---|---|
| GPQA Diamond | 67.2 | 54.8 |
| LiveCodeBench v6 | 37.2 | 69.7 |
| MMLU-Pro | 78 | 67.8 |
| RULER 128K | 93.8 | 55.3 |
| RULER 64K | 95.2 | 67.5 |
| scicode | 36.6 | 24.1 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.